The basic component of a magnetic tunnel junction is a sandwich of two thin ferromagnetic and/or ferrimagnetic layers separated by a very thin insulating layer through which electrons can tunnel. The tunneling current is typically higher when the magnetic moments of the ferromagnetic (F) layers are parallel and lower when the magnetic moments of the two ferromagnetic layers are anti-parallel. The change in conductance for these two magnetic states can be described as a magneto-resistance. Here the tunneling magnetoresistance (TMR) of the MTJ is defined as (RAP−RP)/RP where RP and RAP are the resistance of the MTJ for parallel and anti-parallel alignment of the ferromagnetic layers, respectively. MTJ devices have been proposed as memory cells for nonvolatile solid state memory and as external magnetic field sensors, such as TMR read sensors for heads for magnetic recording systems. For a memory cell application, one of the ferromagnetic layers in the MTJ has its magnetic moment fixed or pinned, so that its magnetic moment is unaffected by the presence of the magnetic fields applied to the device during its operation. The other ferromagnetic layer in the sandwich is the free or sensing layer, whose moment responds to magnetic fields applied during operation of the device. In the quiescent state, in the absence of any applied magnetic field within the memory cell, the sensing layer magnetic moment is designed to be either parallel (P) or anti-parallel (AP) to the magnetic moment of the pinned ferromagnetic layer. For a TMR field sensor for read head applications, one of the ferromagnetic layers has its magnetic moment fixed or pinned so as to be generally perpendicular to the magnetic moment of the free or sensing ferromagnetic layer in the absence of an external magnetic field. The use of an MTJ device as a memory cell in an MRAM array is described in U.S. Pat. No. 5,640,343. The use of an MTJ device as a MR read head has been described in U.S. Pat. Nos. 5,390,061; 5,650,958; 5,729,410 and 5,764,567.
FIG. 1 illustrates a cross-section of a conventional prior-art MTJ device. The MTJ 100 includes a bottom “fixed” ferromagnetic (F) layer 18, an insulating tunnel barrier layer 24, and a top “free” ferromagnetic layer 34. The MTJ 100 has bottom and top electrical leads 12 and 36, respectively, with the bottom lead being formed on a suitable substrate 11, such as a silicon oxide layer. The ferromagnetic layer 18 is called the fixed layer because its magnetic moment is prevented from rotating in the presence of an applied magnetic field in the desired range of interest for the MTJ device, e.g., the magnetic field caused by the write current applied to the memory cell from the read/write circuitry of the MRAM or the magnetic field from the recorded magnetic layer in a magnetic recording disk. The magnetic moment of the ferromagnetic layer 18, whose direction is indicated by the arrow 90 in FIG. 1, can be fixed by forming it from a high coercivity magnetic material or by exchange coupling it to an antiferromagnetic layer 16. The magnetic moment of the free ferromagnetic layer 34 is not fixed, and is thus free to rotate in the presence of an applied magnetic field in the range of interest. In the absence of an applied magnetic field, the moments of the ferromagnetic layers 18 and 34 are aligned generally parallel (or anti-parallel) in an MTJ memory cell (as indicated by the double-headed arrow 80 in FIG. 1) and generally perpendicular in a MTJ magnetoresistive read head. The relative orientation of the magnetic moments of the ferromagnetic layers 18, 34 affects the tunneling current and thus the electrical resistance of the MTJ device. The bottom lead 12, the antiferromagnetic layer 16, and the fixed ferromagnetic layer 18 together may be regarded as constituting the lower electrode 10.
The basic concept of a magnetic tunnel junction was first realized in 1975 (M. Julliére, “Tunneling between ferromagnetic films”, Phys. Lett. 54A, 225 (1975)) although the TMR was very small and observed only at low temperatures and for very small bias voltages. In 1995 significant TMR effects of about 10% were obtained at room temperature in MTJs with Al2O3 tunnel barriers by two different groups (J. S. Moodera et al., “Large magnetoresistance at room temperature in ferromagnetic thin film tunnel junctions”, Phys. Rev. Lett. 74, 3273 (1995); and T. Miyazaki and N. Tezuka, “Giant magnetic tunneling effect in Fe/Al2O3/Fe junction”, J. Magn. Magn. Mat. 139, L231 (1995)). Subsequently, S. S. P. Parkin et al. (“Exchange-biased Magnetic Tunnel Junctions and Application to Non-Volatile Magnetic Random Access Memory”, J. Appl. Phys. 85, 5828 (1999)) obtained effects as large as about 48-50% by optimizing the growth of the Al2O3 tunnel barrier, by optimizing the interface between the Al2O3 tunnel barrier and the ferromagnetic electrodes, and by carefully controlling the magnetic orientation of the ferromagnetic moments using concepts of magnetic engineering, in particular, exchange bias (see U.S. Pat. No. 5,650,958 titled “Magnetic tunnel junctions with controlled magnetic response” to W. J. Gallagher et al.) and an anti-parallel coupled pinned ferromagnetic layer (see U.S. Pat. No. 5,841,692 titled “Magnetic tunnel junction device with antiferromagnetically coupled pinned layer” to W. J. Gallagher et al.).
The magnetoresistance of MTJs using aluminum oxide tunneling barriers is limited to about 50% at room temperature (S. S. P. Parkin et al., “Exchange-biased Magnetic Tunnel Junctions and Application to Non-Volatile Magnetic Random Access Memory”, J. Appl. Phys. 85, 5828 (1999); X.-F. Han et al., “Fabrication of high-magnetoresistance tunnel junctions using Co75Fe25 ferromagnetic electrodes”, Appl. Phys. Lett. 77, 283 (2000)), although there have been reports of TMR values of up to about 58% at room temperature (M. Tsunoda et al., “60% magnetoresistance at room temperature in Co—Fe/Al—O/Co—Fe tunnel junctions oxidized with Kr—O2 plasma”, Appl. Phys. Lett. 80, 3135 (2002)). The detailed structure and composition of the barrier and particularly the structure and composition of the interfaces between the barrier and the ferromagnetic electrodes clearly influences the magnitude of the TMR (as well as the resistance of the junctions). Usually the Al2O3 tunnel barrier is formed by first depositing a thin aluminum layer and then oxidizing this layer either by using an oxygen plasma or by oxidation in oxygen or air. Incomplete or under-oxidation of the barrier may lead to “pin-holes” in the barrier which will usually result in a diminishment of the TMR. On the other hand, over-oxidation of the barrier or excess oxygen within the barrier or at the barrier surface will result in oxidation of the ferromagnetic electrodes which also usually results in decreased TMR. There have been a small number of reports of improved TMR by using special methods of forming Al2O3 tunnel barriers. For example, Tsunoda et al., “60% magnetoresistance at room temperature in Co—Fe/Al—O/Co—Fe tunnel junctions oxidized with Kr—O2 plasma”, Appl. Phys. Lett. 80, 3135 (2002), assert that it is preferred to oxidize the Al layer by using a plasma formed from an inert gas-oxygen mixture where the inert gas is Kr or He. They argue that this method results in an improved barrier layer, because it has previously been shown that SiO2 gate dielectric layers have improved properties (lower number of interface defect states) when formed by oxidation using inert gas—O2 plasmas.
The tunnel magnetoresistance (TMR) of MTJs is also influenced by the ferromagnetic electrode. For electrodes formed from Ni—Fe, Co—Fe or Ni—Fe—Co alloys, it is now generally agreed that there is a surprisingly weak dependence of TMR on the composition of this alloy (S. S. P. Parkin et al., “Exchange-biased Magnetic Tunnel Junctions and Application to Non-Volatile Magnetic Random Access Memory”, J. Appl. Phys. 85, 5828 (1999); D. J. Monsma and S. S. P. Parkin, “Spin polarization of tunneling current from ferromagnet/Al2O3 interfaces using copper-doped aluminum superconducting films”, Appl. Phys. Lett. 77, 720 (2000)), but that the magnitude of the TMR is strongly influenced by the quality of the interface between the ferromagnetic electrode and the Al2O3 tunnel barrier. Once the interface structure is optimized, either by optimizing the growth or by post-growth annealing, for sufficiently thick Al2O3 tunnel barriers which give rise to resistance-area (RA) products exceeding ˜100-500 Ωμm2, TMR values between 40 and 50% can be obtained for almost all of these ferromagnetic alloys. As the tunnel barrier thickness and the corresponding RA value are decreased below this value, it is generally found that the maximum TMR which can be obtained is reduced (see U.S. Pat. No. 6,226,160 titled “Small area magnetic tunnel junction devices with low resistance and high magnetoresistance” to W. J. Gallagher and S. S. P. Parkin, which is hereby incorporated by reference).
For applications of magnetic tunnel junctions for either magnetic recording heads or for non-volatile magnetic memory storage cells, high TMR values are needed for improving the performance of these devices. The speed of operation of the recording head or memory is related to the signal to noise ratio (SNR) provided by the MTJ—higher TMR values will lead to higher SNR values for otherwise the same resistance. Moreover, for memory applications, the larger the TMR, the greater is the variation in resistance of the MTJs from device to device which can be tolerated. Since the resistance of an MTJ depends exponentially on the thickness of the tunneling barrier, small variations in thickness can give rise to large changes in the resistance of the MTJ. Thus high TMR values can be used to mitigate inevitable variations in tunnel barrier thickness from device to device.
The resistance of an MTJ device increases inversely with the area of the device. As the density of memory devices increases in the future, the thickness of the tunnel barrier will have to be reduced (for otherwise the same tunnel barrier material) to maintain an optimal resistance of the MTJ memory cell for matching to electronic circuits. Thus a given variation in thickness of the tunnel barrier (introduced by whatever process is used to fabricate the MTJ) will become an increasingly larger proportion of the reduced tunnel barrier thickness and so will likely give rise to larger variations in the resistance of the MTJ device.
The use of MTJ sensors in magnetic recording read heads has been described in several patents including U.S. Pat. No. 5,729,410 to Fontana and Parkin and U.S. Pat. No. 5,898,548 to Dill et al. The high tunneling magnetoresistance of MTJs makes MTJ sensors attractive candidates for next generation recording read heads beyond current-generation in-plane giant magnetoresistive (GMR) spin-valve sensors, which have limited signal due to the comparatively small values of GMR (˜10-15%). However, one problem with the use of MTJs for read head sensors is their high resistance as compared to metallic GMR sensors. Moreover, the perpendicular current geometry in MTJ read heads means that the sensor resistance increases as the area of the device is reduced for higher capacity disk drives. The higher the resistance of the sensor, the higher is the Johnson noise associated with this resistance. Thus the high TMR of the MTJ device is mitigated by higher noise: what is important is the signal to noise ratio (SNR), which needs to be as large as possible. In order for MTJ sensors to be competitive with in-plane GMR spin valve sensors, resistance-area (A) products as low as a few Ohm-μm2 are needed. Most work on MTJs for recording head sensors has focused on tunnel barriers formed from alumina (Al2O3), which has a large tunnel barrier height. The resistance of an MTJ device increases exponentially with the thickness of the tunnel barrier. Thus, in order to make MTJs with RA values in the range of ˜1 Ohm-μm2, ultra-thin alumina tunnel barriers are required. However, it is very easy to form pin-holes in alumina barriers having the thickness needed to achieve these RA values. Alternative materials have been proposed that have lower tunnel barrier heights such as Ga2O3, as described in IBM's U.S. Pat. No. 6,359,289 to Parkin. However, Ga2O3 exhibits comparatively small TMR values. What is needed are MTJs which have high TMR at RA values lower than what has been achieved with alumina and related barriers. Such MTJ materials are also useful for ultra dense MRAM devices in which the cross-sectional area of the MTJ storage cells is scaled down to very small dimensions, thereby requiring the tunnel barrier to be correspondingly thinner to ensure reasonable resistance values for the MTJ storage cell.
The tunneling current in an MTJ is spin polarized, which means that the electrical current passing from one of the ferromagnetic layers is predominantly composed of electrons of one spin type (spin up or spin down, depending on the orientation of the magnetization of the ferromagnetic layer). The spin polarization P of the current can be inferred from a variety of different measurements. The measurement most relevant to magnetic tunneling is to measure the conductance as a function of bias voltage for junctions formed from a sandwich of the ferromagnetic material of interest and a superconducting counter electrode (R. Meservey and P. M. Tedrow, Phys. Rep. 238, 173 (1994)). These studies show that the spin polarization of the tunnel current measured in this way can be simply related to the TMR close to zero bias voltage as first proposed by Juliere (M. Julliére, Phys. Lett. 54A, 225 (1975)). In such a model P is defined as (n↑−n↓)/(n↑+n↓), where n↑ and n↓ are the density of spin up and spin down states at the ferromagnet/insulator interface. By assuming that the tunnel current is comprised of two independent majority and minority spin currents and that these currents are related to the respective density of states of the majority and minority carriers in the opposing ferromagnetic electrodes, the TMR can be formulated by the relation TMR=(RAP−RP)/RP=2P1P2/(1−P1P2), where RAP and RP are the resistance of the MTJ for anti-parallel and parallel orientation of the ferromagnetic electrodes, respectively, and P1 and P2 are the spin polarization values of the two ferromagnetic electrodes. Experimentally, it is clear that the magnitude of the TMR is extremely sensitive to the nature of the interface between the tunneling barrier and the ferromagnetic electrode. By changing the properties of the interface layer, for example, by inserting very thin layers of non-magnetic metals between the ferromagnet and the insulator layers, the TMR can be dramatically altered. Based on such observations, most experimental data on magnetic tunneling have usually been interpreted by assuming that P is largely determined by the electronic structure of the ferromagnetic interface layer essentially independent of the tunnel barrier electronic structure.
Recently, it has been speculated that the electronic structure of the tunnel barrier may play a more important role than previously realized (W. H. Butler, X.-G. Zhang, T. C. Schulthess et al., Phys. Rev. B 63, 054416 (2001); and P. Mavropoulos, N. Papanikolaou, and P. H. Dederichs, Phys. Rev. Lett. 85, 1088 (2000)). In particular, the primary role of the tunnel barrier was previously assumed to be to determine the evanescent decay length of the electronic wave functions into the tunnel barrier region. Butler et al. and Mavropoulos et al. have argued that the evanescent decay length depends on both the momentum of the electrons transverse to the ferromagnet/insulator interface as well as the Bloch symmetry of these wave functions. Butler et al. have especially considered the case of Fe/MgO/Fe, since it has long been recognized that there is an almost perfect lattice match between the simple cubic structure of the MgO insulator and the body-centered cubic (bcc) structure of Fe for the (100) crystallographic orientation if the lattices are rotated by 45 degrees. Butler et al. find that for the (100) orientation there is a very slow decay into the MgO barrier of majority spin electron states with Δ1 symmetry for small transverse momentum. Thus, for parallel orientation of the ferromagnetic electrodes in an MTJ, these electronic states lead to a very high conductance across the tunnel barrier. Butler et al. calculate that the Fe/MgO/Fe system should exhibit TMR values of hundreds or even thousands of percent. Moreover, Butler et al. calculate that the TMR should have a very strong dependence on MgO tunnel barrier thickness, increasing by orders of magnitude as the MgO thickness is changed by a few atomic layers. Such speculations have led to numerous experimental studies to explore the possibility of high TMR in epitaxial (100) oriented Fe/MgO/Fe tunnel junctions. Note that early work by several groups on MTJs containing polycrystalline MgO tunnel barriers found no evidence for large TMR values.
Some of the first studies of Fe/MgO/Fe MTJs were those by Keavney et al. (D. J. Keavney, E. E. Fullerton, and S. D. Bader, J. Appl. Phys. 81, 795 (1997)) who prepared high quality epitaxial MgO tunnel barriers on Fe single crystal whiskers using molecular beam epitaxy (MBE) growth techniques in which the Fe and MgO layers were deposited by electron beam evaporation. Keavney et al. argued that the MgO incompletely wets the Fe underlayer leading to pin-holes in the MgO layer and, thus, to ferromagnetic coupling of the Fe layers through the MgO layer. The pin-holes through the tunnel barrier electrically shorted the MTJ so that no TMR was observed. These authors concluded that MgO was a very poor choice for a tunneling barrier in MTJs. Later Wulfhekel et al. (W. Wulfhekel, M. Klaua, D. Ullmann, et al., Appl. Phys. Lett. 78, 509 (2001)) prepared high quality epitaxial MgO tunnel barriers on Fe single-crystal whiskers using both molecular beam epitaxy (MBE) and pulsed laser deposition (PLD) growth techniques. This group concluded, by looking at tunneling through the MgO layer using scanning probe microscopy, that a significant portion of the electrical current tunneled through the MgO layer, although there were some local hot spots. The hot spots correspond to some sort of defect or pin-hole. However, although the crystalline quality of these films was very good, this group found no evidence for significant tunneling magnetoresistance in their samples.
The first report of significant tunneling magnetoresistance through epitaxially grown MgO tunnel barriers was by Bowen et al. (M. Bowen, V. Cros, F. Petroff, et al., Appl. Phys. Lett. 79, 1655 (2001)), who reported 60% TMR but at low temperatures (30 K). However, at room temperature this group reported TMR values of only 27%, which is much lower than TMR values that have been obtained with conventional amorphous Al2O3 tunnel barriers. This group studied sandwiches of CoFe/MgO/Fe grown on GaAs(100) with MgO(100) buffer layers by a combination of sputtering (CoFe and Fe) and laser ablation (MgO). The MgO barrier was grown at 400° C., and the Fe and CoFe layers were deposited at room temperature, but the Fe layer was annealed at 400° C. after deposition and prior to deposition of the MgO layer on top of it. The Co50Fe50 layer was the top electrode and was presumably used to allow for different magnetic switching fields for the two ferromagnetic electrodes. This group studied TMR in junctions with very thick Fe and CoFe layers (˜200 Å and 250 Å, respectively)—much too thick for useful applications because of the very large demagnetizing fields that would be produced by these thick layers. This group explored MgO layers in a range of thickness from 20 to 80 Å, and from cross-section transmission electron microscopy studies found good quality crystalline (100) oriented MgO layers for these thicknesses. Based on the predictions by Butler et al. that the TMR for epitaxial MgO tunnel barriers should increase strongly with MgO thickness, Bowen et al. argued that the small TMR values they observed might be increased for thicker MgO tunnel barriers, although they only included tunneling transport data for one MgO layer thickness.
Popova et al. (E. Popova, J. Faure-Vincent, C. Tiusan, et al., Appl. Phys. Lett. 81, 1035 (2002)) have published results on epitaxial 100 oriented Fe/MgO/Fe/Co MTJs deposited by MBE on MgO(100) substrates. This group prepared the Fe layers by evaporation from a Knudsen cell and the Co and MgO layers by electron beam evaporation. The first Fe layer was deposited at room temperature and then annealed at 450° C. after deposition and prior to deposition of the MgO barrier. This group reported modest values of TMR at room temperature of only ˜15% for junctions with 10 Å thick MgO barriers, although the crystalline quality of the structures was very good with smooth and epitaxial Fe and MgO layers. This same group has recently published data on similar structures with thicker tunnel barriers (25 Å thick) in which TMR values of up to 67% were found at room temperature (J. Faure-Vincent, C. Tiusan, E. Jouguelet, et al., Appl. Phys. Lett. 82, 4507 (2003)). They argue that thick MgO tunnel barriers are needed to obtain these higher TMR values, even though the TMR they find is not significantly higher than that which has been observed in MTJs with Al2O3 tunnel barriers. Popova et al. also suggest that the modest TMR values they find, especially when compared to the theoretical predictions of Butler et al., may result from the formation of an FeO layer at the Fe/MgO interface during the deposition of MgO on the lower Fe electrode. The formation of an FeO layer was previously postulated by Meyerheim et al. (H. L. Meyerheim, R. Popescu, J. Kirschner, et al., Phys. Rev. Lett. 87, 076102 (2001)), who found evidence for a such a layer from detailed structural investigations using surface x-ray diffraction of the growth of MgO on single crystal Fe(001) substrates. Recently, X.-G. Zhang, W. H. Butler, and A. Bandyopadhyay (Phys. Rev. B (2003)) have carried out calculations of the TMR for Fe/FeO/MgO/Fe junctions and have found that the presence of an FeO layer substantially reduces the predicted TMR values for this system.
Recently Mitani et al. (S. Mitani, T. Moriyama, and K. Takanashi, J. Appl. Phys. 93, 8041 (2003)) have also attempted to prepare epitaxial Fe/MgO/Fe tunnel junctions by growth on single crystalline MgO(100) substrates. Mitani et al. first deposited an Fe layer (200 Å thick) by electron beam deposition at room temperature with a subsequent post-deposition anneal at 200° C. They then deposited a MgO tunnel barrier by depositing a thin layer of Mg, plasma oxidizing this layer in an Ar—O2 mixture, and then repeating this process several times to create the tunnel barrier. Subsequently they deposited a Co50Fe50 counter electrode on top of the MgO barrier to create the MTJ. Although this group was able to prepare high quality epitaxial tunnel junctions, these junctions showed poor TMR, with TMR values at low temperatures (4.2 K) of only 22.9%. The resistance of the tunnel junctions was found to decrease substantially with temperature, which Mitani et al. argued was due to poor quality MgO tunnel barriers with defects in the barrier, which resulted in hopping conductivity of the tunneling electrons through these defects.
In U.S. Pat. No. 6,392,281, Tsuge discloses a means of forming a magnetic tunnel junction formed from two ferromagnetic layers separated by an oxide tunnel barrier by first depositing the lower ferromagnetic electrode with or without a metal overlayer and then forming an oxide of the metal layer, if present, and the upper portion of the ferromagnetic layer by exposing these layers to ultra-pure oxygen. When the metal layer is not initially present, a metal layer is subsequently formed on top of the ferromagnetic oxide layer and is then subjected to oxidation by pure oxygen gas. Tsuge argues that a subsequent heat treatment will cause oxygen to diffuse from the ferromagnetic oxide layer into the metal oxide, which forms the tunnel barrier, if the heat of formation of the metal oxide is significantly greater than that of the ferromagnetic oxide. However, the devices of Tsuge have considerably lower tunnel magnetoresistance values than those fabricated using other prior art methods of forming the tunnel barrier and magnetic tunnel junction. Evidently, Tsuge does not demonstrate any improvement over the prior art because the heat of formation of the metal oxide versus the formation of the ferromagnetic oxide is not the critical parameter in determining whether oxygen from the ferromagnetic oxide will diffuse away from this layer into the metal oxide layer. Even though the oxygen may be in a lower energy state in the oxide barrier, the metal oxide barrier will be fully oxidized by the process described by Tsuge because of the high heat of formation of the metal layers disclosed therein. Thus, all the oxygen sites in the metal oxide layer will be occupied, which does not allow for the diffusion of oxygen from the ferromagnetic oxide layer. Moreover, the diffusion of oxygen through a ferromagnetic oxide layer is likely to be small, so that the diffusion of oxygen from the ferromagnetic oxide into the metal oxide layer will require extreme conditions of high temperature. In other words, there will be considerable energy barriers to the flow of oxygen from the ferromagnetic oxide layer into the metal oxide layer, even if there are unoccupied oxygen sites in the metal oxide layer, so that oxygen will not diffuse over distances of more than about 1 atomic layer into the metal oxide layer. Thus it is not surprising that the devices formed by Tsuge have very low tunneling magnetoresistance values, because the surface of the ferromagnetic layer will not be free of oxide, in contradiction to the assertions of Tsuge.
In U.S. patent application Ser. No. 10/646,246 to Parkin titled “MgO tunnel barriers and method of formation” and filed Aug. 22, 2003, a method of forming MgO tunnel barriers is described in which magnetic tunnel junctions are prepared with very high tunnel magnetoresistance. However, the band gap of these MgO tunnel barriers may be too high for certain applications.
There is a continuing need for MTJ devices having significantly higher magnetoresistance values than those in the prior art at lower resistance-area product values.